You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
[Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks](https://arxiv.org/abs/1511.06434)
5
+
6
+
### Abstract
7
+
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.
8
+
9
+
## Train
10
+
$ python main.py --mode train \
11
+
--scope [scope name] \
12
+
--name_data [data name] \
13
+
--dir_data [data directory] \
14
+
--dir_log [log directory] \
15
+
--dir_checkpoint [checkpoint directory]
16
+
---
17
+
$ python main.py --mode train \
18
+
--scope dcgan \
19
+
--name_data celeba \
20
+
--dir_data ./datasets \
21
+
--dir_log ./log \
22
+
--dir_checkpoint ./checkpoint
23
+
24
+
* Set **[scope name]** uniquely.
25
+
* Hyperparameters were written to **arg.txt** under the **[log directory]**.
26
+
* To understand hierarchy of directories based on their arguments, see **directories structure** below.
27
+
28
+
## Test
29
+
$ python main.py --mode test \
30
+
--scope [scope name] \
31
+
--name_data [data name] \
32
+
--dir_data [data directory] \
33
+
--dir_log [log directory] \
34
+
--dir_checkpoint [checkpoint directory] \
35
+
--dir_result [result directory]
36
+
---
37
+
$ python main.py --mode test \
38
+
--scope dcgan \
39
+
--name_data celeba \
40
+
--dir_data ./datasets \
41
+
--dir_log ./log \
42
+
--dir_checkpoint ./checkpoints \
43
+
--dir_result ./results
44
+
45
+
* To test using trained network, set **[scope name]** defined in the **train** phase.
46
+
* Generated images are saved in the **images** subfolder along with **[result directory]** folder.
47
+
***index.html** is also generated to display the generated images.
0 commit comments